Quantum Computation Michael A. Nielsen University of Queensland Goals: 1. To eplain the quantum circuit model of computation. 2. To eplain Deutsch s algorithm. 3. To eplain an alternate model of quantum computation based upon measurement.
What does it mean to compute? Church-Turing thesis: An algorithmic process or computation is what we can do on a Turing machine. Deutsch (1985): Can we justify C-T thesis using laws of physics? Quantum mechanics seems to be very hard to simulate on a classical computer. Might it be that computers eploiting quantum mechanics are not efficiently simulatable on a Turing machine? Violation of strong C-T thesis! Might it be that such a computer can solve some problems faster than a probabilistic Turing machine? Candidate universal computer: quantum computer
The Church-Turing-Deutsch principle Church-Turing-Deutsch principle: Any physical process can be efficiently simulated on a quantum computer. Research problem: Derive (or refute) the Church- Turing-Deutsch principle, starting from the laws of physics.
Models of quantum computation There are many models of quantum computation. Historically, the first was the quantum Turing machine, based on classical Turing machines. A more convenient model is the quantum circuit model. The quantum circuit model is mathematically equivalent to the quantum Turing machine model, but, so far, human intuition has worked better in the quantum circuit model. There are also many other interesting alternate models of quantum computation!
Quantum circuit model Unit: bit Classical Quantum Unit: qubit,,..., n 1 2 1. Prepare n-bit input 2. 1- and 2-bit logic gates 3. Readout value of bits 1. Prepare n-qubit input in the computational basis. 2. Unitary 1- and 2-qubit quantum logic gates 3. Readout partial information about qubits Eternal control by a classical computer.
Single-qubit quantum logic gates Pauli gates 0 1 0 i 1 0 X = ; Y ; Z 1 0 = = i 0 0 1 Hadamard gate H 0 + 1 0 1 1 1 1 H 0 = ; H 1 = ; H = 2 2 2 1 1 Phase gate P P 0 = 0 ; P 1 = i 1 1 0 P = 0 i P P = 2 P = Z Z
Controlled-not gate Control Target c t c t c 1 0 0 0 0 1 0 0 0 0 0 1 0 0 1 0 U CNOT is the case when U=X Controlled-phase gate 1 0 Z = 0 1 Z ct ct, ( 1) ct, Z = Z Symmetry makes the controlled-phase gate more natural for implementation! = X H Z H Eercise: Show that HZH = X.
Toffoli gate Control qubit 1 c 1 c1 Control qubit 2 c2 c2 Target qubit t t c 1 c 2 Worked Eercise: Show that all permutation matrices are unitary. Use this to show that any classical reversible gate has a corresponding unitary quantum gate. Challenge eercise: Show that the Toffoli gate can be built up from controlled-not and single-qubit gates. Cf. the classical case: it is not possible to build up a Toffoli gate from reversible one- and two-bit gates.
How to compute classical functions on quantum computers Use the quantum analogue of classical reversible computation. The quantum NAND The quantum fanout y y 1 0 1 y Classical circuit Quantum circuit f( ) f f( ) Uf 0 m g
Removing garbage on quantum computers z z f( ) 0 m U f f( ) g U f 0 m Canonical form: z z f( ) Eample: z z parity( ) Given an easy to compute classical function, there is a routine procedure we can go through to translate the classical circuit into a quantum circuit computing the canonical form.
Eample: Deutsch s problem { } { } Given a black bo computing a function f : 0,1 0,1 Our task is to determine whether f is constant or Classically we need to evaluate both f(0) and f(1). balanced? Quantumly we need only use the black bo for f () once! Classical black bo f z z f( ) Quantum black bo U f z z f( )
Putting information in the phase f( ) = 0: f( ) = 1: 0 1 2 ( 0 1 ) ( 0 1 ) U f ( 0 1 ) ( 1 0 ) = ( 0 1 ) f ( ) ( 0 1 ) ( 1) ( 0 1 ) ( 1) f( )
Quantum algorithm for Deutsch s problem 0 0 1 2 0 0 + 1 H f ( ) + ( ) (0) f (1) 1 0 1 1 Uf f (0) f (1) ( 1) ( 0 1 ) + ( 1) ( 0 1 ) + H Quantum parallelism ( ) ( ) ( ) ( ) f (0) f (1) f (0) f (1) = 1 + 1 0 + 1 1 1 f constant all amplitude in 0. f balanced all amplitude in 1. Research problem: What makes quantum computers powerful?
Universality in the quantum circuit model Classically, any function f() can be computed using just nand and fanout; we say those operations are universal for classical computation. Suppose U is an arbitrary unitary transformation on n qubits. Then U can be composed from controlled-not gates and single-qubit quantum gates. Just as in the classical case, a counting argument can be used to show that there are unitaries U that take eponentially many gates to implement. Research problem: Eplicitly construct a class U n of unitary operations taking eponentially many gates to implement.
Summary of the quantum circuit model Inpu: t An n-bit string,, representing an instance of some problem. Eample: is a number to be factored. m Initial state: 0, where m is some computable function of n. Circuit: A circuit of single-qubit and controlled-not gates is applied to the qubits. The sequence of gates applied is under the control of an eternal classical computer, and may depend upon the problem instance. Readout: Some fied subset of the qubits is measured in the computational basis at the end of the computation, and the output constitutes the solution to the problem. Eample: For a decision problem, just the first qubit would be read out, to indicate "yes" or " no". QP: The class of decision problems soluble by a quantum circuit of polynomial size, with polynomial classical overhead.
Quantum compleity classes How does QP compare with P? BQP: The class of decision problems for which there is a polynomial quantum circuit which outputs the correct answer ( yes or no ) with probability at least ¾. BPP: The analogous classical compleity class. Research problem: Prove that BQP is strictly larger than BPP. Research problem: What is the relationship of BQP to NP? What is known: BPP BQP PSPACE
When will quantum computers be built?
Alternate models for quantum computation Standard model: prepare a computational basis state, then do a sequence of one- and two-qubit unitary gates, then measure in the computational basis. Research problem: Find alternate models of quantum computation. Research problem: Study the relative power of the alternate models. Can we find one that is physically realistic and more powerful than the standard model? Research problem: Even if the alternate models are no more powerful than the standard model, can we use them to stimulate new approaches to implementations, to error-correction, to algorithms ( high-level programming languages ), or to quantum computational compleity?
Overview: Alternate models for quantum computation Topological quantum computer: One creates pairs of quasiparticles in a lattice, moves those pairs around the lattice, and then brings the pair together to annihilate. This results in a unitary operation being implemented on the state of the lattice, an operation that depends only on the topology of the path traversed by the quasiparticles! Quantum computation via entanglement and singlequbit measurements: One first creates a particular, fied entangled state of a large lattice of qubits. The computation is then performed by doing a sequence of single-qubit measurements.
Overview: Alternate models for quantum computation Quantum computation as equation-solving: It can be shown that quantum computation is actually equivalent to counting the number of solutions to certain sets of quadratic equations (modulo 8)! Quantum computation via measurement alone: A quantum computation can be done simply by a sequence of two-qubit measurements. (No unitary dynamics required, ecept quantum memory!) Further reading on the last model: D. W. Leung, http://.lanl.gov/abs/quant-ph/0111122
Can we build a programmable quantum computer? U ψ U fied array U ψ PU, ψ No-programming theore m: Unitary operators U1,..., U which are distinct, even up to global phase factors, require orthogonal programs U,..., U. Challenge eercise: Prove the no-programming theorem. 1 n n
A stochastic programmable quantum computer ψ Bell Measurement j = 0,1,2,3 U U I U ( ) 00 + 11 2 Uσ ψ j
Why it works ψ Bell Measurement j = 0,1,2,3 00 + 11 2 U U Uσ j ψ U I U ( ) 00 + 11 2 σ j ψ
How to do single-qubit gates using measurements alone ψ Bell Measurement j = 0,1,2,3 00U + 11 2 k U σ k U σ k ( ) 00 + 11 Uσ σ ψ 1 Uk I Uσ k With probability 2 4, j = k, and the gate succeeds. k j
Coping with failure Action was Uσ σ, j k - a known unitary error. k j Now attempt to apply the gate UUσσ ( k j ) to the qubit, using a similar procedure based on measurements alone. Successful with probability 1 4, otherwise repeat. ( ) Failure probability ε can be achieved with O log 1 repetitions. ε
How to do the controlled-not ψ Bell 2 Measurement j, k = 0,1,2,3 U lm ( )( ) m j k U σ σ σ σ ψ l lm ( ) U I Uσ σ l 1 16 m Bell With probability, j = l, k = m, and the gate succeeds. 2
Discussion Measurement is now recognized as a powerful tool in many schemes for the implementation of quantum computation. Research problem: Is there a practical variant of this scheme? Research problem: What sets of measurement are sufficient to do universal quantum computation? Research problem: Later in the week I will talk about attempts to quantify the power of different entangled states. Can a similar quantitative theory of the power of quantum measurements be developed?